Land is one of the most prominent geo-features in marine navigation for passage planning and execution. Land generalization for nautical charting, and particularly that of smaller landmasses (islands), is a task performed manually by cartographers based on available specifications and their own experience. Existing methods for automated island selection utilize the Voronoi diagram or weighted buffers to calculate island density and remove islands in dense regions. Their main deficiency is that they do not consider other related chart features, while they may also incorrectly represent the density and distribution of islands and the detection of isolated ones in the process. As such, islands that pose navigational significance and would be retained by a professional cartographer may be removed by the algorithm and vice versa. These drawbacks restrict their use as they present difficulties in achieving the product requirements. This paper presents an island generalization method that combines the benefits of Voronoi and buffer approaches, while it incorporates relevant other chart feature classes to respect topology relations and, with island shape recognition, to categorize islands on a hierarchy level for making generalization decisions. This innovative approach enhances the effectiveness and accurate presentation of islands on nautical charts through automated generalization.

Current nautical chart generalization methods are notably labor intensive, requiring significant levels of human intervention to compile, update, and maintain chart products. The ideal situation would be a fully automated solution for generating nautical charts seamlessly from a comprehensive database, on demand, at the appropriate scale, at the point of use, and respecting the product constraints. However, regardless of the various research efforts and advancements in technology, including those involving AI, nautical chart generalization tasks are still performed manually, or semi-manually, where a likelihood of human error is expected. This manuscript presents a research effort toward automated chart compilation through scales. Nautical chart generalization guidelines are extracted, categorized, and translated into machine readable rules, utilized by a multi-agent model to perform the generalization of the source data to the target scale with no topological violations. This is illustrated in three testbeds for the most important ENC feature classes. While topology is maintained, the model utilizes readily available algorithms that, generally, compromise safety. Therefore, a custom validation tool detects safety violations for user intervention. The model has been made flexible to incorporate algorithms that align with application constraints, especially safety, as they become available.